Frequency analysis shows results that are way off

[velle]

I recorded my voice saying a word, "my", monotomely in my normal voice. I want to find out the pitch, and I thought I could do so with Audacity. I marked a small interval (300ms) of the monotome "aaaa" part. I used Analyze -> Plot Spectrum. It gave a peak on 278Hz.

This is surprising because I am a male with a voice in the deep end and according to Wikipedia (http://en.wikipedia.org/wiki/Voice_frequency):
Typical male voice: 85-180Hz
Typical female voice: 165-255Hz

I then generated a sine wave (within Audacity) with 440Hz and analyzed that, but it shows a peak on 416Hz.

These results are useless! What could they ever be used for? Or did I misunderstand something?

Sugggestions on how to get the pitch of the recroding are very welcome.

Sincerely, Velle

Audacity 2.0.2, Windows 7

[steve]

"Plot Spectrum" uses FFT analysis. The accuracy of FFT analysis depends on the FFT "Size" parameter. For the most accurate frequency reading, set the "Size" parameter as large as possible for the selection. If the selection is very short you may not be able to use the maximum size, in which case decrease the size one step at a time until you reach the largest "Size" that can be used with the selection, or make a larger selection. The largest "Size" that can be used with a 300 ms selection when the sample rate is 44100, is 8192.

See here in the manual for more details: http://manual.audacityteam.org/o/man/an ... l#spectrum

[velle]

Thanks That helped. The analysis of the sine wave gives exactly 440 with a large window. I have two dominating peaks, and many smaller peaks, when analyzing my voice:

126Hz: -7.8Hz
232Hz: -8.6Hz

I'm not sure what the second I get these, so close, with almost identical amplitudes, but I will simply pick the first one as my "result" since it is in the expected range.

[Robert J. H.]

Those values are not so close as it may seem. They are in fact nearly an octave apart (B2 to Bb3). In other words, it is the jump from "Some..." to "Over..." in "Somewhere over the rainbow".
The first value is the fundamental frequency, whereas the second is most likely the first formant. You can observe how this value changes when you're singing different vocals, although tthe pitch (of the fundamental) stays the same.

[kozikowski]

The mess of overtones and frequencies kills people trying to filter out words or instruments by "looking for the note and deleting that frequency." Those extra spikes are the reason you don't sound like Harry Connick Jr and neither do I even though we're singing the same notes.

Robert may sound like HCJ. Jury's out.

Koz